[MPlayer-dev-eng] Seeking beta testers for libnut!

[Michael Niedermayer]

Hi

On Fri, Jan 27, 2006 at 05:29:02PM +0200, Ivan Kalvachev wrote:
> 2006/1/27, Oded Shimon <ods15 at ods15.dyndns.org>:
> > Well, it works for me. :)
> >
> > I'm asking everyone to test libnut as much as possible... tell me about
> > assert failure, crashes, any other weird problems, bad seeks.
> >
> > Don't bother feeding libnut even slightly broken avi files, it will not
> > tolerate them.
> >
> > As for broken NUT files, well, truncated files will work just fine, but
> > other invalid files probably won't work well. Tell me if it works, don't
> > bother telling me if it doesn't.
> >
> > Most importantly, check seeking, seeking should be percise and never result
> > in a position higher than requested pts.
> >
> > If you see the error "NUT error: Cannot seek to that position.", when you
> > made a small seek and the video gets a bunch of mpeg-4 artifacts, don't
> > bother telling me, it's a bug in mplayer, not in libnut. libnut couldn't
> > find a keyframe closer to position you requested then the position you are
> > at right now, so it failed. No way to inform mplayer about a failed seek,
> > and it craps out the video...
> That's your bug not mplayer.
> 
> I had stated before that you must implement both backward (aka
> precise) and forward seeking.
> 
> The backward seeking is when you prefer keyframe before requested pts.
> Forward is when you prefer frame after the requested pts (in case you
> cannot get exact pts). This also applies when you want to seek
> backward or forward, you choose corresponding backward or forward
> seeking so you will get away from the current position.
> We don't want WMP behaviour where seeking 5 seconds ahead takes 4.5 seconds.

that wouldnt be so bad, but trying to seek 5 seconds forward and ending up 0.5
seconds ahead would be ...

[...]

-- 
Michael